Behjoee, Alireza (2025) Forex Trend Prediction: Can Machine Learning Algorithms Forecast Short Timeframe Movements? Bachelor thesis, Data Science and Society (DSS).
|
PDF
Thesis---Alireza-Behjoee.pdf Download (749kB) | Preview |
Abstract
The Foreign Exchange (FOREX) market is the largest and most liquid financial market in the world. Its worldwide 24-hour accessibility, coupled with its dynamic nature, makes it very attractive to traders and investors. Therefore, numerous deep learning algorithms have been developed to forecast market trends, but they face challenges such as overfitting, long training times, and limited interpretability. Additionally, most prior research has focused on longer timeframes, such as daily, which may not be optimal for capturing the frequent fluctuations in Forex prices. This study attempts to bridge these gaps by developing machine learning models using the 13 most commonly used technical indicators, in addition to OHLCV (Open, High, Low, Close, V olume) data. The models are trained on shorter timeframes, specifically the 5-minute, 15-minute, 30-minute, 1-hour, and 4-hour intervals, for the EUR/USD currency pair and the XAU/USD commodity pair. Then, they are evaluated not only based on accuracy but also through profitability analysis in order to assess their performance under real-world trading conditions and how much net profit they could generate. Overall, our proposed model achieves an accuracy of 82.10% and a profit of 9.2% for XAU/USD on the 5-minute timeframe, based on backtesting from April 25, 2025, to May 1, 2025.
| Item Type: | Thesis (Bachelor) |
|---|---|
| Name supervisor: | Haleem, N. |
| Date Deposited: | 05 Aug 2025 10:15 |
| Last Modified: | 05 Aug 2025 10:15 |
| URI: | https://campus-fryslan.studenttheses.ub.rug.nl/id/eprint/751 |
Actions (login required)
 |
View Item |